ES cluster

(trelo) #1

Hi, experts! I'm newbie in ES and I need your advice.
I've an app which is hosting in two different sites(A and B). In site A I've a single master ES node which is handling all requests from both A and B site(multiple backends in every site), and which is "Single Point Of Failure". What is the best way to add Failover for ES to this app and avoid "brain split" if something wrong with network between A and B site? Just give me direction.

(David Pilato) #2

It depends on your case.

If you need to be able to index in both sites and replicate on the 2 sites immediately.

If not, let say you index in site A. You can run a snapshot every x minutes from site A and restore in site B.
If you need more "real time" search, instead of indexing in elasticsearch site A, you can send the operation to something like Kafka which you set to be distributed in site A and B, then read with Logstash for example and index in both sites the same data.

In the future, we will have probably something like cross data center replication but it's not yet there. With that you'll probably be able to do the later automatically...

(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.